First, thanks for the suggestions. I have updated my config file and you could find it here:
http://purgatory.spnet.net/~karavelov/slapd.conf.clean
But the problem I have desctribed still persist. You could find a log here:
http://purgatory.spnet.net/~karavelov/faillog
the logging is set to: trace stats stats2 pcache (I believe that 0x1000 is pcache)
It consists of 3 parts:
1. First query for "justillusion.net", it is not in the cache. It hits the SQL database and is reported to be stored in dbd cache 2. Some seconds later the same query fails 3. An analogous query succeed
I could cut more form the log file if somebody needs it.
I do not think that the fails depend on the data being stored (no NULLs). Some different records get corrupted and I do not see nothing common that could be the cause of the corruption. Just in differend times, differend records get corrupted.
This is slapd-2.4.17 + patch from head (back-sql/search.c), libdbd-4.6
Thanks in advance